1 Whoever Saves One Life Saves the World: Confronting the Challenge of Pseudoinefficacy Daniel Västfjäll Decision Research and Cognitive Psychology, Linköping University Paul Slovic Decision Research, 1201 Oak Street, Suite 200, Eugene, Oregon, and Department of Psychology, University of Oregon, Marcus Mayorga Decision Research and Department of Psychology, University of Oregon Abstract In a great many situations where we are asked to aid persons whose lives are endangered, we are not able to help everyone. What do we then do? In a series of experiments, we first demonstrate that donors, in general, become demotivated by information about children who cannot be helped. We find that negative affect from the children not helped decreases the warm glow associated with aiding the children who can be helped. This demotivation may be a form of pseudoinefficacy that is nonrational. We should not be deterred from helping whomever we can because there are others we are not able to help. Second, we show that people react in two ways to such requests. Some feel less good about helping those they can help and they help less. Others feel badly because of those out of reach and they become even more motivated to help whomever they can. We discuss the need to better understand these two different reactions and we suggest strategies to reduce the demotivating effects of pseudoinefficacy. Key words: pseudoinefficacy, singularity effect, prosocial behavior, psychic numbing, compassion
2 Acknowledgments This material is based upon work supported by the Hewlett Foundation, and by the National Science Foundation under Grant Nos. SES , , and Manuscript under review, April 2014 From the movie Schindler s List: Oskar Schindler: I could have got more out. I could have got more. I didn t know. If I d just... I could have got more. Itzhak Stern: Oskar, there are eleven hundred people who are alive because of you. Look at them. Oskar Schindler: If I d made more money... I threw away so much money. You have no idea. If I d just... Itzhak Stern: There will be generations because of what you did. Oskar Schindler: I didn t do enough! Itzhak Stern: You did so much. [Schindler looks at his car] Oskar Schindler: This car. Goeth would have bought this car. Why did I keep the car? Ten people right there. Ten people. Ten more people. [removing Nazi pin from lapel] Oskar Schindler: This pin. Two people. This is gold. Two more people. He would have given me two for it, at least one. One more person... and I didn t! And I... I didn t! In a climactic scene toward the end of the movie Schindler s List, factory owner Oskar Schindler is given a gift of a ring by the 1,100 Jewish workers whose lives he had saved by sheltering them from the Nazis. On the inner band of the ring is an inscription from the Talmud: Whoever saves one life saves the world entire. As he slips the ring on his finger, his thoughts turn to other lives he might have saved but did not, and he breaks into tears because I could
3 have got more out. The research to be described in this paper suggests that if Schindler had thought earlier of lives he was not protecting from death, he might not have protected those 1,100 he did save. What motivates us to help others whose lives are endangered? More specifically, what motivates us to help in certain situations, while in others we turn away? The answer to this question depends on the extent to which we value potential or actual losses of lives. Normatively, the scope or magnitude of a disaster or crisis should be the main carrier of value and motivation to act. But, descriptively, our actions are sometimes insensitive to, or even demotivated by, increasing numbers of people at risk (Slovic, 2007). For example, a single identified victim often evokes stronger feelings and greater willingness to help than an unidentified single victim or a group of victims, identified or not. The preference for helping a single identified victim over a group of victims is known as the singularity effect (Kogut & Ritov, 2005b). In other circumstances, decision makers appear to be constructing their life-saving preferences on the basis of contextual information that may not be normatively justifiable. In a study by Jenni and Lowenstein (1997), participants evaluated a program saving 2 lives annually more favorably when those 2 lives were half of a population of 4 at risk, than when they were a much smaller percentage of 1,700 at risk. Slovic, Finucane, Peters, and MacGregor (2002) termed this type of effect proportion dominance (see also Bartels, 2006), and argued that affective feelings play a central role in this phenomenon. In the present article, we examine another contextual factor that may not be normatively justified: information about lives we cannot save may demotivate us from saving those we can. Pseudoinefficacy Decisions are strongly motivated by perceived efficacy (Cryder, Loewenstein, & Scheines, 2013, Erlandsson, Björklund, & Bäckström, 2013). Inefficacy, real or perceived, shrivels response, even among those who have the desire and the means to protect and improve lives. It is tragic, indeed, when efficacy goes unrecognized and vital aid that could be provided is withheld due to an illusion of ineffectiveness that we have named pseudoinefficacy. In reviewing what appeared to be unrelated findings from two earlier studies of life-saving decisions, we uncovered a curious connection that motivated the present research. These prior studies asked people to provide clean water to aid people facing death from disease (Fetherstonhaugh, Slovic, Johnson, & Friedrich, 1997) or to provide money to protect a child from
4 starvation (Small, Loewenstein, & Slovic, 2007). Fetherstonhaugh et al. found that people were less likely to send clean water that could save 4,500 lives in a refugee camp when the number of people in the camp was large (250,000) than when it was small (11,000). Small et al. found that the money donated to a seven-year-old African child facing starvation decreased dramatically when the donor was made aware that the child was one of millions needing food aid. The findings from these two studies may have broad implications for prosocial or humanitarian behavior in light of the insights of Andreoni (1990), who contended that we help others not only because they need our help but because we anticipate and experience the warm glow of good feeling associated with giving aid. Subsequent empirical studies have supported this contention (e.g., Dunn, Aknin, & Norton, 2008). We hypothesize that knowledge of those out of reach (more in the large refugee camp and millions of starving people in Africa) may have triggered negative feelings that countered the good feelings anticipated from giving aid, thus demotivating action. A related explanation is that, compared to the large numbers of persons out of reach, the prospective aid created a sense of inefficacy, that is, a drop-in-the-bucket effect (Bartels & Burnett, 2011). Although the results from these studies by Fetherstonhaugh et al. and Small et al. may appear at first glance to reflect inefficacy, this is not really inefficacy, because the donor can actually help some people (from 1 to 4,500). Instead, it is a form of pseudoinefficacy that is nonrational. We should not be deterred from helping one person, or 4,500, just because there are others we cannot help. We propose that the demotivating effects in these studies, all of which involved large numbers of unidentified children who could or could not be helped, are one specific form of pseudoinefficacy. Kahneman (2011) summarizes the extensive research documenting the differential effects of fast vs. slow thinking (see also Greene, 2013). In a related vein, Haidt (2001) argues that moral intuitions (fast thinking) precede and often dominate moral reasoning (slow thinking). We propose that the findings of Fetherstonhaugh et al. (1997) and Small et al. (2007) reflect relatively slow or reasoned pseudoinefficacy that arises from more complex thoughts involving calculations of proportions or drop-in-the-bucket imagery, likely causing feelings of despair or hopelessness. In the present article we extend these findings to situations involving what we call fast or intuitive pseudoinefficacy. We propose that fast pseudoinefficacy is linked to virtually immediate dampening of warm glow by
5 negative feelings, perhaps of sadness or unhappiness, in situations with small numbers of identified people in need and small numbers unable to be helped. We describe a series of studies designed to examine fast pseudoinefficacy and clarify the psychological processes that contribute to it. To test our hypothesis that, even when the numbers of affected individuals are small, awareness of those not helped reduces the warm glow arising from doing good things, we employ a paradigm involving donations to one or more starving children identified by name, age, photo, and so on. We systematically vary the number of children who can be helped and the number who cannot. For instance, in a between-groups design, those in Group 1 are asked: How good do you feel about donating to help Child A? Those in Group 2 are asked to help Child A but not Child B, who unfortunately cannot be helped. In this case, we predict that warm glow associated with helping Child A would be lower for Group 2 and would decrease further if more excluded children were highlighted. We also predict that donations of money will decrease as warm glow is diminished. Overview Experiments In Study 1, an initial test of pseudoinefficacy was conducted wherein participants rated their feelings and indicated how much money they would donate to either one child (who would be helped for certain) or one of two children (1 child cannot be helped, but uncertain which child). In Study 2, we employed two experimental conditions where 1 child is helped and the remaining 5 children cannot be helped. In the uncertain condition participants were not told which child would be helped. In the certain condition participants were explicitly made aware of which children they could and could not help. Study 3 introduced ratings of warm glow. It included more variations of the number of children helped or not helped (always specifying which children could be helped, and which could not be helped). Studies 4a and 4b employed within-subjects designs where participants rated warm glow as a function of the number of children helped. Studies 5a and 5b tested whether including irrelevant neutral visual distractors or affect-inducing non-children pictures would produce dampening of warm glow for helping a child. Studies 6a and 6b extended the previous studies of warm glow to include monetary donations. Study 1: Initial demonstration of pseudoinefficacy This initial study was designed to extend the large-number paradigms used in previous studies (Fetherstonhaugh et al., 1997; Small et al., 2007) by
6 introducing a paradigm where participants are asked to donate to a either single child or one out of two children. Method and design Ninety-four undergraduates (48 males) at Göteborg University, Sweden with a mean age of 27.3 (SD 5.1) participated in this study. A procedure devised by Small et al. (2007) was used in which participants, after completing an unrelated survey, received seven Swedish ten-kronor coins, a blank envelope, a questionnaire including the three response measures, and a charity request letter. The experimenter instructed participants to first read the charity request letter carefully, then place their donations (if any) in the envelope. Next, participants were asked to complete the questionnaire and return both the letter and questionnaire in the sealed envelope. The letter informed the participant of the opportunity to donate any of their just-earned 70 kronors to the organization Save the Children. Participants were randomly allocated to one of the single or uncertain conditions. First, there were two single-child conditions: a description and picture of either a seven-year-old girl, Rokia (n = 21) or a nine year-old boy, Moussa (n = 22). Participants were instructed that Any money that you donate will go to Rokia [Moussa]. Rokia [Moussa] is desperately poor, and faces a threat of severe hunger or even starvation. Her [His] life will be changed for the better as a result of your financial gift. With your support, and the support of other caring sponsors, Save the Children will work with Rokia s [Moussa s] family and other members of the community to help feed her [him], provide her [him] with education, as well as basic medical care and hygiene education. In the uncertain condition (n = 51), participants received a similar description but with pictures and stories of both Rokia and Moussa. Participants were instructed that their donations would go to Rokia or Moussa. Three response measures were used: 1. Willingness to donate. Participants could circle any number between 0 and 70 Swedish Crowns (SEK) in 10-crown increments. 2. Affect. Participants rated how do you feel about donating to Rokia/Moussa/the child? on a scale ranging from Negative ( 1) to Positive (+5).
7 3. Perceived probability that the donation would make a real difference (1 5 scale anchored by Not at all likely to Very likely). Results and discussion Table 1 shows means comparing the uncertain condition to the average of the two single conditions. Independent t-tests showed that donations were significantly higher in the single conditions, t(92) = 2.12, p <.05. Similarly, affect ratings were more positive in the single conditions, t(92) = 2.46, p <.05.). The perceived probability did not differ between conditions, t(92) = 1.76, p =.08. Table 1 Willingness to donate, affect, and probability for uncertain and single conditions Single child Uncertain One of two children (n = 43) (n = 51) Donations (SEK) Affect Probability Although donors may consider their contributions to be a drop in the bucket in some circumstances, such thinking is unlikely to underlie the results of this study, given that half of the two children at risk could be helped. Because the instructions in the uncertain condition explicitly stated that one of the two children would be helped, we believe that participants considered the one child not helped in the uncertain conditions and that this reduced affect and donations. Study 2: Comparing uncertain and certain conditions where children are not being helped It is possible that results from Study 1 were affected by the uncertainty regarding which child would be helped (Gneezy, List, & Wu, 2006). Previous research has found that uncertainty elicits negative affect (Kurtz, Wilson, & Gilbert, 2007). In Study 2, we controlled for this possibility by directly contrasting an uncertain condition (one of n will be helped, but uncertain which
8 one), and certain condition (this child will be helped for sure; these others will not be helped). Method and design One hundred and four undergraduates at Göteborg University, Sweden (65 females) with a mean age of 24.1 participated in Study 2. We collected ratings of warm glow and affect in a between-groups design where participants either saw one child who could be helped (n = 37) or one of two scenarios where one child could be helped and six could not be helped. In the uncertain condition (n = 33) participants were not told which one of the seven children could be helped. In the certain condition (n = 34), the child who could be helped was identified. In the single-child condition, an identified child with a photo and a name was presented. Participants were instructed that: This is a picture of Nayani. Her living conditions are very bad and she needs your help. Without your help Nayani will likely not survive. You can help Nayani by donating money. Below are a number of questions about your thoughts and feelings about helping Nayani. The certain condition included seven pictures, each depicting a named child. Participants were instructed that: This is picture of some children. Their living conditions are very bad and they need your help. Without your help these children will likely not survive. You can help one of these children by donating money. The child you can help is located to the left. Her name is Nayani. The presentation included identifying information about Nayani and a color photo, while photos of the six other children were faded and printed in black and white. In the uncertain condition participants saw pictures of all seven children (all were named and presented in color) and were instructed that: This is picture of some children. Their living conditions are very bad and they need your help. Without your help these children will likely not survive. You can help one of these children by donating money. Below are a number of questions about your thoughts and feelings about helping one of the children.
9 Participants in all conditions rated their feelings using three scales anchored by 0 (Not at all) to 7 (Very much): 1. Warm glow: If I donated money, I would experience a warm glow feeling. 2. Positive affect: I have positive feelings when I think about Nayani/the child. 3. Compassion: I have compassionate feelings for Nayani/the child. Participants were then debriefed and thanked for their participation. Results and discussion ANOVAs on the dependent measures yielded significant effects on all scales. Warm-glow, positive affect, and compassion ratings were significantly lower in the certain condition than in the single-child condition. The uncertain condition was not different from the certain condition on any of the three scales (see Table 2). Table 2 Mean ratings of warm glow, positive affect, and compassion for the three conditions in Study 2 One child Uncertain Certain F p Warm glow 3.3a 2.9a,b 2.5b 4.46 <.02 Positive affect 2.7a 2.2a,b 2.0b 3.28 <.05 Compassion 4.1a 3.4b 3.2b 4.95 <.01 Note. Subscripts not shared are significantly different at p <.05. Together, these findings show that information about children not helped dampens affect for the child who could be helped. However, affect in the certain condition was not significantly lower than in the uncertain condition, suggesting that participants in both these conditions were demotivated by the children not helped. These findings suggest that the negative feelings associated with awareness of the children not being helped dampens feelings for the child that could be helped.
10 Study 3: Varying numbers of children helped or not helped In Study 3, we further extended the certain paradigm to test how variations of the number of children helped or not helped influence warm glow. Method and design Five hundred forty-three U.S. participants from a nation-wide sample (mean age 38, 56% female) completed an online version of the experiment. Study 3 directly examined the effect on warm glow of the number of children helped or not helped. Participants again saw pictures of children. In the first scenario seen by a participant we varied the number helped or not helped in a between-subjects design: 1/0 (helped/not helped), 1/1, 1/6, 2/0, and 2/1 (see Figure 1 for an example of one condition). However, after seeing the initial scenario, participants in each condition saw all other scenarios in mixed orders, thus allowing us to also do an analysis of ratings within-subjects. The instructions read: In this survey you will be presented with children in need of help. We will ask you to think about warm glow, a positive feeling that you may experience when you do something good for someone. Take a moment to think about one situation from your own life when you experienced this feeling. Next we will ask you to consider the warm glow you would expect to feel if you donated to help children in need. Participants were then shown one of the conditions with a picture and names of the child(ren). For example, in the 1/6 condition they were instructed that These are seven children in need of aid. The child on the left, Ofelo, is facing starvation and is in immediate need of food. The six children on the right are facing ill health from water-borne diseases and are in immediate need of clean water and medicines. Suppose that you are now given the opportunity to donate money to a trusted aid organization to help Ofelo (the child on the left). Unfortunately, you can only help Ofelo and not the other children, for whom aid may not be available. The warm-glow scale was an open-ended response anchored by 0 (No warm glow) to 100 (Very strong warm glow). Figure 1. Example of stimulus materials for one condition.
11 Results and discussion Means for both within- and between-subject comparisons are shown in Table 3. The overall between-subjects ANOVA was significant: F(1,540) = 2.39, p <.05. As can be seen in Table 3, for both between and within- subjects designs, warm glow associated with helping one child decreased linearly as the number of children not helped increased from zero not helped, to one not helped, and then to six not helped. Two children helped elicited warm glow similar to one child helped. A within-subject ANOVA (collapsing across different orders) also showed a significant condition effect, F(1,540) = 35.57, p < Comparing two helped and two helped/one not helped in the between subjects design, a non-significant increase in warm glow occurred. However, within subjects, two helped/one not helped exhibited a statistically significant decrease in warm glow compared to two helped. Table 3 Mean warm-glow ratings for both between- and withinsubject comparisons (Study 3) Helped/Not helped 1/0 1/1 1/6 2/0 2/1 (n = 137) (n = 100) (n = 110) (n = 86) (n = 80) Between-subjects 54.9 a 50.8 b 47.1 c 55.8 a 60.4 a Within-subjects (n = 502) 53.6 a 49.6 b 45.4 c 53.7 a 51.4 b Note. Means not sharing a subscript are different at p <.05. Table entries are warm-glow ratings on a scale. Study 3 thus replicated the basic pseudoinefficacy effect and, further, showed that warm-glow ratings were sensitive to the number of children not helped. With the increasing number not helped, warm-glow feelings were further dampened. The effects were very similar for both within- and betweensubjects comparisons. Studies 4a and 4b: Testing robustness of the effect in within-subjects designs Previous research (Hsee & Zhang, 2010, Kogut & Ritov, 2005a) has documented that in joint evaluation participants typically adjust their responses (so that they are the same) when they realize that their judgment is about the same object or stimulus (in this case the warm glow for a single
12 child). The within-subjects results in Study 3, however, suggest that the pseudoinefficacy effect may resist this type of judgmental correction. In Study 4, we further tested this resistance by assessing warm glow in a within-subjects design with a fixed order of presentation of several different donation opportunities. Study 4a was conducted as a classroom exercise and Study 4b was a replication using an individualized computer survey in a laboratory setting. Method and design Study 4a One hundred and forty-three students in a college classroom at the University of Oregon participated in this study. About two thirds were women. Their mean age was about 20. The number of children who could be helped was systematically varied in the following fixed order: six helped/one not helped, two helped/one not helped, one helped/one not helped, one helped. The general instructions were as follows: Here are some questions about children in Africa who live in poverty. I will ask you to consider helping these children by donating money to a respected aid organization and then answer a number of questions about your thoughts and feelings. In particular I would like you to think about warm glow a positive feeling that you may experience when you do something good for someone. Have you experienced this? Take a moment and think about one situation from your own life when you experienced this feeling. Next I will ask you to consider the warm glow you expect to feel if you donated to help children in need. The first rating was for warm glow expected if one donated money to help six children (pictured, with names) but could not donate to a seventh, pictured and named child, as follows: These are seven children in need of aid. Suppose that you are given the opportunity to donate money to a trusted aid organization to help the six children to the left (Nelson, Sueva, Moussa, Mutaka, Jallo, Maluf). Unfortunately you can only help these six children and not Okeke, for whom aid may not be available. Rate the warm glow you expect to feel
13 if you donated money to help these six children (Nelson, Sueva, Moussa, Mutaka, Jallo, and Maluf). Participants rated their warm glow by pressing a button on an audiencevoting system (participants were instructed that the response options represented the following intervals on a scale of warm glow: 1 = 0 20, 2 = 21 40, 3 = 41 60, 4 = 61 80, 5 = ). Our data-collection method did not track individual answers, so only means are reported for this study. Study 4b Forty-eight University of Oregon undergraduates (mean age 20.5 years; 75% female) participated in a lab study. The methodology was similar to Study 4a, except participants rated the pictures in a computer survey. They responded they responded using the same five-category scale of warm-glow used in Study 4a. Responses were tracked within-subjects, allowing statistical tests to be performed. Results and discussion In both Studies 4a and 4b, warm glow decreased as the number of children who could be helped decreased (see Figure 2). A within-subjects ANOVA showed a significant condition effect F(1, 46) = 11.80, p <.01 for Study 4b. Importantly, the critical difference between one child helped/one not helped and one helped showed that the mean warm glow was substantially higher in the one-child condition. In Study 4b, this difference was significant in a Bonferroni post-hoc test (p <.01). Figure 2. Mean warm-glow ratings in within-subjects designs (Studies 4a and 4b). Together, Studies 4a and 4b extended our earlier results in two ways: In contrast to the earlier studies, we kept the number of children not helped constant (one), and varied the number helped. We found that participants experienced less warm glow as the number of children helped decreased, a finding that is consistent with proportion dominance (Fetherstonhaugh et al., 1997). However, the single child helped received the highest warm-glow rating consistent with the singularity effect (Kogut & Ritov, 2005a). Importantly, when the single child was paired with one other child not helped, that child then received the lowest warm-glow rating. It is notable that this effect occurred even with a fixed order where the single child appeared last and, arguably, participants would likely recognize that in both the one helped/one not helped and the one-helped scenarios they may help only one child and thus
14 need not change their warm-glow ratings. We believe this is a demonstration of the robustness and pervasiveness of dampening of good feelings for the child one can help when paired with one or more children one cannot help. Studies 5a and 5b: Testing an alternative explanation and a possible mechanism Studies 5a and 5b had three goals; (1) provide evidence that the children not helped induce negative affect that dampens the positive affect for the child helped, (2) examine whether pseudoinefficacy could be produced by simply introducing distracting stimuli, and (3) examine whether the role of negative affect could be isolated by incidentally manipulate negative affect. In Study 5a we contrasted the effect on ratings of warm glow of children who cannot be helped with the effect of visual distractors. If the dampening of warm glow observed in the previous studies was due to mere distraction caused by attending to the children not helped, we would also expect to find reduced warm-glow ratings in conditions with visual distractors other than children. In Study 5b we tested the hypothesis that a responsible mechanism for pseudoinefficacy is negative affect associated with the children not helped. We also compared the effect of children who cannot be helped with the effect of other visual stimuli that induce negative affect. We expected that other sorts of irrelevant pictures that induce negative affect would also reduce warm glow, consistent with our main explanation for the effect. Study 5a: Visual distractors Method and design One hundred and forty-eight undergraduates at Göteborg University, Sweden (mean age 32, 68% female) participated in an online survey. To test the effect on warm glow of visual distractors, we compared a condition (n = 54) where one child could be helped and six could not be helped with a condition (n = 44) where six shapes were substituted for the six children not helped (see Figure 3). In addition, a single-child condition was included (n = 44). In each of the three conditions, participants were given the same general instructions to think about warm glow as were given in our prior studies and they were asked to rate the warm glow they expected to feel for the single child they could help, using a 0 (No warm glow) to 100 (Very strong warm glow) response scale. Participants in the one/six children condition also rated how they felt when they viewed the children not helped (on a 2 to 2 scale from very
15 bad to very good). Participants in the shapes condition similarly rated how bad or good they felt when they viewed the shapes on a four-point scale ranging from Very bad ( 2) to Very good (+2). Figure 3. Child and shapes used in the visual distraction experiment in Study 5a. Results and discussion As can be seen in Figure 4, mean warm glow was significantly lower in the one child-helped/six-children-not-helped condition than in the single-child condition (t(96) = 2.35, p <.03), replicating our previous findings. The one child/six shapes condition did not differ from the single-child condition (t(96) = 0.91, p =.92), suggesting that visual distraction is not the cause of the observed pseudoinefficacy. Consistent with a negative-affect explanation, participants experienced significantly more negative affect (M =.28) when viewing the children not helped than when viewing the shapes (M =.31); (t(86) = 3.81, p = <.001). Further, the correlation across participants between warm-glow ratings and the valence ratings of the children not helped was positive and significant, r =.32, p <.05, while the correlation between warm glow and valence ratings of the shapes was r =.00, ns. Figure 4. Mean warm-glow ratings across the different conditions in the distractor study. Study 5b: Manipulating negative affect In Study 5b we sought to determine the role of negative affect in dampening warm glow. The design was similar to Study 5a but, instead of affect-neutral shapes, pictures selected to induce strong negative affect were shown. Method and design One hundred-four undergraduates at Göteborg University, Sweden with a mean age of 23 years (61% female) participated in a web-based survey. To test the effect of negative affect, we compared a condition (n = 26) where one child could be helped and six could not with a condition (n = 43) in which six negative affect-inducing pictures were substituted for the six children not helped (see Figure 5). The pictures were selected from the International Affective Pictures System (IAPS; Bradley & Lang, 2000). As in Study 5a, a single-child condition was included (n = 35). In addition to rating warm glow for the child, participants in the one-plus-six children condition rated how bad or good they felt when viewing the children not helped ( 2 to + 2 scale).